بازسازي حسگري فشرده MRI مبتني بر شبكه هاي متخاصم مولد عميق با استفاده از U-net بهبود يافته

چكيده انگليسي :

Compressive Sensing MRI Reconstruction Using improved U net based on Deep Generative Adversarial Networks GANs Seyed Amir Mousavi amir mousavi1@ec iut ac ir Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 IranDegree Master of Science Language FarsiSupervisor Mohammad Reza Ahmadzadeh Assoc Prof Ahmadzadeh@cc iut ac ir AbstractMagnetic resonance imaging MRI as a non invasive imaging is able to produce three dimensional detailedanatomical images without the use of damaging radiation and with excellent visualization of the anatomicalstructure MRI is a time consuming imaging technique Image quality may be reduced due to spontaneousor non spontaneous movements of the patient Several imaging techniques like parallel imaging have beensuggested to enhance imaging speed Compressive Sensing MRI CS MRI violates the Nyquist Shannon sam pling rate and utilizes the sparsity of MR images to reconstruct MR images with under sampled k space data Prior studies in CS MRI have employed orthogonal transforms such as wavelets and recent methods have em ployed dictionary learning for adaptive transforms CS MRI is an efficient way to decrease the scan time ofMR imaging However the computational costs are usually expensive and the CS reconstruction process istime consuming In addition CS MRI methods are based on constant transform bases or shallow dictionaries which limits modeling capacity Deep Learning is a novel orientation in Machine Learning and Artificial Intel ligence investigation It has already been shown that CNNs work better than sparsity based approaches in termsof both image quality and reconstruction speed In this thesis a novel method based on very deep convolutionalneural networks CNNs for reconstruction of MR images is proposed using Generative Adversarial Networks GANs In this model a Generative and Discriminator networks designed with improved ResNet architecture The Generative network is based on U net and Discriminator is a classifier network which improved blocksis used in both of them Using improved architecture has led to deepening of Generative and discriminatornetworks reduction in aliasing artifacts more accurate reconstruction of edges and better reconstruction of tis sues To achieve better reconstruction adversarial loss pixel wise cost and perceptual loss pre trained deepVGG network are combined Two comparisons have been made with the latest studies in this field The firstcomparison has been made with DLMRI which is a well known conventional method of CS MRI to recon struct magnetic resonance images The second comparison has been made with deep learning method namedDAGAN Compared to DLMRI and DAGAN methods it has been demonstrated that the proposed methodoutperforms the conventional methods and deep learning based approaches Assessment is made on severaldatasets such as the brain heart and prostate and the proposed method leads to a better reconstruction in de tails of the images Reconstruction of brain data with a radial mask of 30 in the proposed method has beenimproved on basis of the SSIM criteria up to 0 98 Also image reconstruction time is about 20 ms on GPU thatis much smaller than the state of the art CS MRI methods KeywordsDeep Learning Generative Adversarial Networks GANs MRI Compressive Sensing

موسوي، امير

بازسازي حسگري فشرده MRI مبتني بر شبكه هاي متخاصم مولد عميق با استفاده از U-net بهبود يافته

يادگيري عميق , شبكه هاي متخاصم مولد , تصويربرداري تشديد مغناطيسي , حسگري فشرده